Peter Schuller wrote:
> It would be embarressing if I caused this problem myself by
> misunderstanding wal_archiving. My understanding has been that once
> wal_archive gets called, no one ever cares what happens with the file
> except if I want to do PITR (since the whole point is to offload it
> somewhere or similar).
>
> I'll go and re-read the documentation on this immediately. If this is
> the problem, I do apologies for the noise and people's time.
Yeah, this is the problem. The archive command must not delete the
file it is copying. It is supposed to only copy it. The file must
continue to be on the server until a checkpoint is executed.
In fact you'd have a problem on recovery after any crash, not just
vacuum full.
I think it's the documentation's fault that this is not a lot more
prominent.
--
Alvaro Herrera http://www.CommandPrompt.com/
PostgreSQL Replication, Consulting, Custom Development, 24x7 support
